Semantic Understanding Of Mangrove Remote Sensing Images Based On Decision Rules

Remote sensing image classification and detection based on computer technology is an important means to monitor the dynamic changes of land cover.Aiming at the complexity of mangrove environment monitoring in Dongzhai Port,Hainan,a semantic understanding method based on the kernel joint sparse representation classifier(KJSR)and decision rules is proposed.The purpose is to classify the types of mangroves and the ground cover plants around them,and divide them into two categories according to the growth of mangroves: good growth and poor growth.Finally,the decision rules for the scene are given.First,this paper studies the classification performance of the kernel joint sparse representation classifier for remote sensing images with different resolutions.Experiments show that in the high-resolution image classification task,the overall classification accuracy reaches 99.32%,which is at least 1.01% higher than the typical classifier.In the low-resolution image classification task,the overall classification accuracy reaches 99.6%,which is at least1.6% higher than the typical classifier.The sparse representation classifier can effectively improve the recognition rate of sample features in the model training process,and has a good learning effect and classification performance.Secondly,a method for semantic understanding of remote sensing images based on decision rules is proposed,which uses the results of KJSR as decision features,and divides mangrove and non-mangrove areas.Then select the best segmentation attribute according to the maximum information gain criterion,divide the two types of mangroves,and extract the semantic rules for the scene division.The results show that this method can effectively identify feature types.